JP2021135945A - System and method for assisting creation of game script - Google Patents

Abstract

To provide a system capable of assisting creation of a game script in narrative game development.SOLUTION: A learning support device 10 for assisting creation of a game script includes: a data dividing section 21 that classifies a created game script by character and stores the created game script for each character in a storage device; a data pre-processing section 22 that converts control data, which is classifiable into a plurality of types depending on function and contained in a created game script created in advance, into a control descriptive text being natural language data for each type of the control data, and that creates processed script text containing descriptive text and control descriptive text corresponding to the descriptive text; and a learning section 23 that generates a trained model for each type of control data by causing a natural language trained model in which grammatical structures and relationships between sentences related to natural language are learned in advance to the processed script text for each type of control data.SELECTED DRAWING: Figure 2

Description

The present invention relates to a system and a method for supporting the creation of a game script.

Conventionally, various games have been released, including games executed on small portable electronic devices such as smartphones. For example, in a game with a story such as an RPG, there is a game in which the story progresses while switching the standing picture or dialogue of the character. In such a narrative game, the story progresses according to the items described in the "game script", which is a scenario text consisting of character lines with a command sequence that controls the progress of the story such as screen transitions. do.

JP-A-2018-169715

Conventionally, when creating a game script in a game having a story as described above, it is necessary to input all manually, and the game script creation process has been one of the cost factors in game production. In such a situation, in the development of a narrative game, there is a need for a system for supporting the creation of a game script that can reduce manual involvement without degrading the quality. For example, if at least a part of a game script such as input of a command sequence responsible for progress control can be automatically created, manual involvement can be reduced without degrading the quality. Various efforts have been made by the applicant for automation in game development with a story, for example, Patent Document 1 provides a translation support system capable of reducing translation costs by using a machine translation system. It is disclosed.

The present invention has been made to solve such a problem, and an object of the present invention is to provide a system capable of supporting the creation of a game script in the development of a narrative game.

The system as one aspect of the present invention is
[1] A system for supporting the creation of a game script including natural language data representing a description in a game associated with the content of the game and control data for controlling the game.
Data preprocessing unit that converts the control data included in the pre-created game script into a control description that is natural language data, and creates a processed script statement that includes the description and the control description corresponding to the description. When,
A learning unit that generates a trained model by training a processed script sentence in a natural language pre-learned model in which the grammatical structure related to the natural language and the relationship between sentences are learned in advance.
It is characterized by including.

Further, the system as one aspect of the present invention is
[2] Control data can be classified into multiple types according to the function.
The data preprocessing unit converts the control data included in the pre-created game script for each type of control data into a control description which is natural language data, and the description and the control description corresponding to the description. Create a processed script statement containing the statement and
The learning unit trains a processed script sentence in a natural language pre-learned model in which the grammatical structure related to the natural language and the relationship between sentences are learned in advance for each type of control data. This is the system according to [1], which generates a trained model of.

Moreover, the system as one aspect of the present invention is
[3] The data preprocessing unit includes a plurality of data preprocessing units, and one data preprocessing unit controls and describes control data corresponding to one type of control data included in the created game script. By converting to a statement and creating a processed script statement, the data preprocessing unit creates a processed script statement for each type of control data.
The learning unit is the system according to [2], which includes a plurality of learning units corresponding to each of the plurality of data preprocessing units.

Moreover, the system as one aspect of the present invention is
[4] The learning unit uses the processed script statement as training data to fine-tune the natural language pre-learned model to generate a trained model, whichever is one of [1] to [3]. The system described.

Moreover, the system as one aspect of the present invention is
[5] The processed script statement further includes a description and a randomly selected control description.
The learning unit causes the natural language pre-learned model to learn the explanatory text included in the processed script statement and the control explanatory text corresponding to the explanatory text as correct data, and the explanatory text and randomly selected control. The system according to any one of [1] to [4], which trains an explanatory text as incorrect data to generate a trained model.

Moreover, the system as one aspect of the present invention is
[6] The game script is matrix-formatted data or structured data, each of which includes a plurality of identifiers corresponding to individual scenes of the game, and natural language data and control data associated with the identifiers. The system according to any one of 1] to [5].

Moreover, the system as one aspect of the present invention is
[7] The game script further includes natural language data representing the character name associated with the identifier.
It further includes a data division unit that classifies created game scripts by character and stores the created game scripts for each character.
The data preprocessing unit converts the control data included in the created game script into a control explanation for each character, and creates an explanation and a processed script including the control explanation corresponding to the explanation.
The learning unit is the system according to [6], which generates a trained model for each character by training a processed script statement for each character in a natural language pre-learned model.

Moreover, the system as one aspect of the present invention is
[8] The data preprocessing unit converts the control data into the control description based on the conversion information indicating the correspondence between the control data and the control description.
The system
An input reception unit that accepts input of explanations in the game,
The input reception unit accepted the input using the trained model generated by training the processed script sentence in the natural language pre-trained model in which the grammatical structure and the relationship between sentences related to the natural language were trained in advance. It is an inference unit that infers a control explanation from the explanation, and the processed script statement is a natural language data created from the explanation included in the created game script created in advance and the control data corresponding to the explanation. The inference part, which contains a certain control description,
The system according to any one of [1] to [7], which includes a data post-processing unit that creates control data from a control explanatory text inferred by the inference unit based on the conversion information.

Moreover, the method as one aspect of the present invention is
[9] A method of generating a trained model for supporting the creation of a game script including natural language data representing a description in a game associated with the content of the game and control data for controlling the game. hand,
A step of converting the control data included in the pre-created game script into a control description which is natural language data, and creating a processed script statement including the description and the control description corresponding to the description.
A step to generate a trained model by training a processed script sentence in a natural language pre-learned model in which the grammatical structure related to the natural language and the relationship between sentences are learned in advance.
It is characterized by including.

Moreover, the system as one aspect of the present invention is
[10] A system for supporting the creation of a game script including natural language data representing a description in a game associated with the content of the game and control data for controlling the game.
An input reception unit that accepts input of explanations in the game,
The input reception unit accepted the input using the trained model generated by training the processed script sentence in the natural language pre-trained model in which the grammatical structure and the relationship between sentences related to the natural language were trained in advance. It is an inference unit that infers a control explanation from the explanation, and the processed script statement is a natural language data created from the explanation included in the created game script created in advance and the control data corresponding to the explanation. The inference part, which contains a certain control description,
It is characterized by including.

Moreover, the system as one aspect of the present invention is
[11] Further includes a data post-processing unit that creates control data from the control explanation inferred by the inference unit based on the conversion information indicating the correspondence between the control data and the control explanation.

Moreover, the method as one aspect of the present invention is
[12] A method for supporting the creation of a game script including natural language data representing a description in a game and control data for controlling the game.
Steps to accept input of description in the game,
Control from the explanatory text that accepts the input using the trained model generated by training the processed script sentence in the natural language pre-trained model in which the grammatical structure and the relationship between sentences related to the natural language are trained in advance. The processed script statement is a step of inferring the explanatory statement, and the processed script statement is a control explanatory statement that is natural language data created from the explanatory statement included in the created game script created in advance and the control data corresponding to the explanatory statement. Including, steps and
It is characterized by including.

A program as one aspect of the present invention is characterized in that a computer is made to perform each step of the method described in [9] or [12].

According to the present invention, it is possible to support the creation of a game script in the development of a game with a story.

It is a block diagram which shows the hardware structure of the learning support apparatus of 1st Embodiment of this invention. It is a functional block diagram of the learning support device of one Embodiment of this invention. It is a figure which shows an example of a game script. It is a figure which shows an example of a game screen 40. It is a figure which shows an example of the created game script which is related to one type of control data. It is a figure which shows an example of the processed script statement created by the data preprocessing unit from the created game script shown in FIG. It is a figure which shows an example of the conversion table which shows the correspondence between the command data of command1 and the control description. It is a figure which shows an example of the processed script statement created by the data preprocessing unit from the created game script shown in FIG. It is a figure which shows the correspondence relationship of the character string included in an image path and an emotion, and is the figure which shows an example of the conversion table which shows the correspondence between an image path and a control explanation. It is a flowchart which shows the generation process of the trained model of the learning support apparatus of one Embodiment of this invention. It is a block diagram which shows the hardware structure of the generation support apparatus of the 2nd Embodiment of this invention. It is a functional block diagram of the generation support apparatus of one Embodiment of this invention. It is a figure which shows an example of the estimation of the control explanatory text by an inference unit. It is a flowchart which shows the automatic creation process of a part of the game script of the generation support apparatus of one Embodiment of this invention.

Hereinafter, a device (system) for supporting the creation of a game script according to an embodiment of the present invention will be described with reference to the drawings. Unless otherwise specified, the same reference numerals in the respective drawings indicate the same or corresponding parts, and for convenience of explanation, the vertical and horizontal scales of the drawings may be expressed differently from the actual ones. Further, for convenience of explanation, an unnecessarily detailed description may be omitted. For example, detailed explanations of already well-known matters and duplicate explanations about substantially the same configuration may be omitted.

The learning support device 10 of the first embodiment of the present invention is a device that generates a trained model in order to support the creation of a game script. The generation support device 50 of the second embodiment of the present invention is a device that automatically creates at least a part of a game script by using the created learned model in order to support the creation of the game script. The learning support device 10 is one embodiment of a learning support system composed of a plurality of devices and the like, but in the following embodiments, it will be described as one device for convenience of explanation. The same applies to the generation support device 50.

The game according to the embodiment of the present invention is a narrative game in which the story progresses while switching the standing picture and dialogue of the character according to the matters described in the game script by the game developer. For example, a game program executed in an electronic device such as a smartphone executes a process for a game by reading a game script.

The game script includes natural language data and control data associated with the content of the game, for example, with identifiers corresponding to individual scenes of the game. For example, a game script is based on the minimum information necessary for the progress of the game, such as lines, characters, and effects described in the scenario text, and natural language data such as character names and lines, and character behavior, standing pictures, and background images. , Background music, sound effects, and other control data are input.

Natural language data is data representing natural language, and includes explanatory texts such as character lines and explanations of situations. The control data is data other than natural language data, and includes command data for controlling the game and data related to the image to be displayed on the screen such as a path for storing the image. For example, command data is data for controlling the progress of a story such as character movements and screen transitions, and is described in a format in which the intended game control is executed when it is read by a game program. NS. The control data can be classified into a plurality of types according to the function and contents of the control data, such as command data and data related to an image. The command data of one type of the control data may be further classified according to the function and contents of the command data.

FIG. 3 is a diagram showing an example of a game script. In this embodiment, the game script is data in matrix format. The row elements are those identified by the ID 31, and the column elements are the data associated with the ID 31. The column elements associated with the ID 31 include the column elements of the character name 32, the description 33, the image path 34, and the command data 35. Note that the row elements and column elements can be interchanged.

Each of the IDs 31, i.e. each of the rows, corresponds to an individual scene of the game, and each of the column elements is associated with an individual scene of the game by being associated with ID 31. Preferably, ID 31 comprises a number, which indicates a chronological order. ID31 is an example of an identifier.

The character name 32 is natural language data representing the names of the characters that appear. In the game script shown in FIG. 3, the character name 32 describes only the elements of the character1_name column corresponding to one character. In this case, the character corresponds to the speaker character, but the game script can increase or decrease the elements of the column corresponding to the character name 32, and the character name 32 corresponding to one or more characters other than the speaker character. Can contain elements of the column of. In one preferred example, the character name 32 is a character ID that can identify the character instead of the character's name. In this case, the character ID has a one-to-one correspondence with the character name which is the natural language data, and since the character name can be specified, it can be regarded as the natural language data. By reading the game script, the game program can refer to the table in which the character ID and the character name are associated with each other.

The explanatory text 33 is natural language data representing at least one of the character's lines or the explanation of the situation.

The image path 34 indicates a path in which the character image is stored, and is one of the control data. The game program acquires the character image by accessing the storage destination. For example, when the character name 32 associated with one ID 31 is not specified, for example, when "null" is described in the character name 32, the explanatory text 33 associated with the one ID 31 is natural language data representing a description of the situation. It becomes.

The command data 35 represents a command for controlling the game corresponding to one scene of the game, and is one of the control data. For example, the command data 35 is a command that specifies a display method of the character image acquired by the image path 34 associated with the same ID 31. In the game script shown in FIG. 3, the command data 35 describes only the elements of the column of command1, but the command data 35 may include elements of a plurality of columns of command data 35 such as command1 and command2. can. In the present embodiment, command1 is command data related to control of movement of the character and the background. In the present embodiment, the command data 35 of the elements in one column corresponds to the command data 35 of one type. For example, one type of command data 35 is command1, and command1 can include a plurality of types of command data such as "fadein" and "fadeout". However, the command data 35 of the elements of the plurality of columns may be configured to correspond to the command data 35 of one type. The game script shown in FIG. 3 can also include elements in columns other than the command data 35.

In one example, the command data 35 includes elements of a column of command data 35 corresponding to each of the characters indicated by the character name 32, and includes elements of a column of command data 35 corresponding to the entire game that does not correspond to a character. Can be done. In another example, the elements of one column of command data 35 are command data related to the control of the movement of the character and the background, and the elements of the other column of command data 35 are facial expressions and expressions that match the lines of the character. This is a command for specifying a standing picture that is posing.

FIG. 4 is a diagram showing an example of the game screen 40. The game screen 40 includes a standing picture 41, a character name 42, and an explanatory text 43. The standing picture 41 is a picture of a character appearing in the game, and is displayed according to the movement of the speaker and facial expressions. The character name 42 is the name of the speaker character. Explanatory text 43 is a line or a description of the situation.

In one example, when the game script is read, the game program acquires the character image corresponding to the standing picture 41 via the image path 34, and executes a process of displaying the acquired character image in the game screen 40. .. In this case, preferably, the game program executes a process of displaying the character image according to the control method described in the command data 35. In one example, the game program executes a process of acquiring the character name 32 and the description 33 when the game script is read. When the character name 32 associated with the same ID 31 as the description 33 is specified, the description 33 is a dialogue, and the game program displays the character name 32 and the character name 42 corresponding to the description 33 in the game screen 40. And the process of displaying the explanation 43 is executed. When the character name 32 associated with the same ID 31 as the description 33 is not specified, the description 33 is a description of the situation, and the game program displays the description 43 corresponding to the description 33 in the game screen 40. Execute the process to be performed.

FIG. 1 is a block diagram showing a hardware configuration of the learning support device 10 according to the first embodiment of the present invention. The learning support device 10 includes a processor 11, an input device 12, a display device 13, a storage device 14, and a communication device 15. Each of these components is connected by a bus 16. It is assumed that an interface is interposed between the bus 16 and each component device as needed. The learning support device 10 includes a configuration similar to that of a general server, PC, or the like.

The processor 11 controls the operation of the entire learning support device 10. For example, the processor 11 is a CPU. The processor 11 executes various processes by reading and executing a program or data stored in the storage device 14. The processor 11 may be composed of a plurality of processors.

The input device 12 is a user interface that receives input from the user to the learning support device 10, and is, for example, a touch panel, a touch pad, or a button. The display device 13 is a display that displays an application screen or the like to the user of the learning support device 10 under the control of the processor 11.

The storage device 14 includes a main storage device and an auxiliary storage device. The main storage device is a semiconductor memory such as a RAM. The RAM is a volatile storage medium capable of reading and writing information at high speed, and is used as a storage area and a work area when the processor 11 processes information. The main storage device may include a ROM, which is a read-only non-volatile storage medium. The auxiliary storage device stores various programs and data used by the processor 11 when executing each program. The auxiliary storage device may be any non-volatile storage or non-volatile memory as long as it can store information, and may be removable.

The storage device 14 stores a created game script, which is a game script created in advance. The created game script is a game script for causing the learning support device 10 to learn. The storage device 14 stores a character image that is referred to when the created game script is read.

The communication device 15 is a wireless LAN module capable of exchanging data with another computer such as a user terminal or a server via a network. The communication device 15 can be another wireless communication device such as a Bluetooth (registered trademark) module, or can be a wired communication device such as an Ethernet (registered trademark) module or a USB interface.

FIG. 2 is a functional block diagram of the learning support device 10 according to the embodiment of the present invention. The learning support device 10 includes a data dividing unit 21, a data preprocessing unit 22, and a learning unit 23. In the present embodiment, these functions are realized by executing the program by the processor 11. For example, the program to be executed is a program stored in the storage device 14 or received via the communication device 15. In this way, since various functions are realized by reading the program, a part or all of one part (function) may be possessed by another part. However, these functions may be realized by hardware by configuring an electronic circuit or the like for realizing a part or all of each function.

First, the overall operation of the learning support device 10 will be described. The learning support device 10 is trained by the learning unit 23 using the processed script statement created by the data preprocessing unit 22 for each type of control data, and generates a learned model for each type of control data. When the learning support device 10 trains the created game script related to the control data in which one type of control data depends on the individuality of the character, the data division unit 21 classifies the created game script for each character and for each character. Create a created game script. In this case, the learning support device 10 is trained by the learning unit 23 using the processed script statements created for each character by the data preprocessing unit 22. When the learning support device 10 learns the created game script related to the control data in which one type of control data does not depend on the individuality of the character, the data division unit 21 does not classify the created game script.

The data division unit 21 classifies the created game scripts for each character and stores the created game scripts for each character in the storage device 14. The classification for each character is a classification for each character name 32 corresponding to the speaker character. In one preferred example, the data divider 21 identifies which speaker character the data associated with the ID 31 in the created game script, i.e., the data in each row, is for which speaker character, and each row for each speaker character. Create a created game script for each speaker character by aggregating the data of. The data division unit 21 stores the created game script created for each character in the storage device 14 for each character. The data associated with the ID 31, that is, the data in each line can be identified as which speaker character the data is related to by referring to the character name. The data division unit 21 does not classify each character depending on the type of control data.

The data preprocessing unit 22 creates a processed script statement by converting the control data included in the created game script stored in the storage device 14 into a control explanation statement which is natural language data for each type of control data. The processed script statement is natural language data including the explanatory statement 33 and the control explanatory statement corresponding to the explanatory statement 33, and is created based on the created game script. When the created game script is divided by the data dividing unit 21, the data preprocessing unit 22 converts the control data included in the created game script stored in the storage device 14 for each character into a control description, and the character. Create a processed script statement for each.

In the present embodiment, the data preprocessing unit 22 includes a plurality of data preprocessing units 122, each of which has a function of the data preprocessing unit 22. Each of the data preprocessing units 122 corresponds to one of the types of control data included in the created game script. Each of the data preprocessing units 122 converts the control data corresponding to one type of the control data included in the created game script into a control description to create a processed script statement. In this way, the data preprocessing unit 22 creates a processed script statement for each type of control data. However, the data preprocessing unit 22 may be realized by one data preprocessing unit, another software module, or the like. For example, when a character that has been "fade in" is to be "fade out", the data preprocessing unit 122 converts "fade in" and "fade out" in the command data included in the created game script into control explanations and processes them. Create a completed script statement.

In one example, the data preprocessing unit 122 converts the control data of the created game script into a control description, and associates or combines the description 33 and the control description associated with the same ID 31 to process the data. Create a script statement. The processed script statement is created one by one in the row data associated with the ID 31 based on the created game script. The processed script statement indicates the entire natural language data created from the created game script, but when it indicates the natural language data associated with ID 31 of one of the data created from the created game script. There is also.

In one preferred example, the data preprocessing unit 122 is shown in Table 1 using one row of data associated with the ID 31 and the natural language data of the character name 32 and description 33 of the created game script. Create one descriptive text 33'such as. The description 33'is natural language data associated with or combined with the control description.

Here, the operations of the data preprocessing unit 22 and the learning unit 23 when the type of control data to be learned is command 1 (command data 35) will be described. FIG. 5 is a diagram showing an example of a created game script related to one type of control data, and FIG. 6 is a processed script statement created by the data preprocessing unit 122 from the created game script shown in FIG. It is a figure which shows an example. FIG. 5 shows a game script related to command1. Here, command1 is control data relating to the control of the movement of the character and the background. In this case, since the control data does not depend on the individuality of the character, the data division unit 21 does not divide the created game script for each character for one type of control data including command1.

The data preprocessing unit 122 processes the natural language data described in the column elements of the character name 32 and the explanatory text 33 shown in FIG. 5 into the natural language data described in the column elements of the explanatory text 33'shown in FIG. .. As shown in FIG. 6, each line of the processed script statement is composed of natural language data. For convenience of explanation, in the processed script statement shown in FIG. 6, the description 33'and the control description 35' are associated with the ID for the purpose of corresponding to the ID 31 of the game script, but the data of each line is managed. Anything that is possible will do.

The "marker indicating that the part controls the movement of the character" shown in Table 1 corresponds to the natural language "character is" in FIG. The natural language "NoChar" in FIG. 6 is an abbreviation for "NoCharacter" and means that the character name is not specified. When the character name 32 is "null", the data preprocessing unit 122 creates the natural language data of the explanatory text 33'so that the "character" after the "marker" is "NoChar". Therefore, when the "front character" and the "speaker character" are not specified, "characters are described as NoChar, NoChar".

In FIG. 5, the scene where the ID 31 is "1" is the first scene, the "character in the foreground" is absent, and only the "catalina" of the "speaker character" exists. In this case, the data preprocessing unit 122 created the explanatory text 33'so that the description after the "marker" of the ID 31 is "characters are NoChar, Catalina". Further, in the scene where the ID 31 is "2" in FIG. 5, "Bee" is designated as the character name 32 as the "speaker character", and "catalina" of the "speaker character" of the previous scene is designated. No. In this case, the data preprocessing unit 122 created the explanatory text 33'so that the "speaker character" is "bee" and the "front character" is "catalina".

The storage device 14 stores control data conversion information indicating the correspondence between the control data and the control description for each type of control data. The control data conversion information is, for example, information in which each of the contents of the command data 35 is associated with each of the control explanations when the control data is the command data 35, and is stored in a format such as a conversion table. The data preprocessing unit 122 converts the control data into a control description based on the control data conversion information.

FIG. 7 is a diagram showing an example of a conversion table showing the correspondence between the command data 35 of command1 and the control description 35'. The data preprocessing unit 122 converts the command data 35 shown in FIG. 5 into the control explanatory text 35'shown in FIG. 6 by using the conversion table shown in FIG. 7. For example, the data preprocessing unit 122 converts the command data "fadein" into the control description "(character) has appeared" by using the conversion table shown in FIG. 7.

The learning unit 23 trains a processed script sentence in a natural language pre-learned model in which the grammatical structure related to the natural language and the relationship between sentences are learned in advance for each type of control data, so that each type of control data Generate a trained model of. When the data preprocessing unit 22 creates a processed script statement for each character, the learning unit 23 learns the processed script statement created for each character for each type of control data in a natural language pre-learned model. By letting it generate a trained model for each character.

The natural language learned model is a learning model created by learning a large amount of natural language sentences in advance by using learning of grammatical structure and learning of relationships between sentences. For learning the grammatical structure, for example, in order to learn the structure of the sentence "My dog is happy", (1) word masking "My dog is [MASK]", (2) random permutation of words "My dog is apple". , (3) No word operation It means to learn the three patterns of "My dog is hairy". Learning the relationship between sentences is performed, for example, when there are two consecutive sentence sets to be learned, the original two sentence sets (correct answer set) and the randomly selected sentence set (incorrect answer set). ) And half are created, and it means learning whether or not the sentences are related as a binary classification problem.

The natural language learned model is stored in another device different from the learning support device 10, and the learning support device 10 communicates with the other device via the communication device 15 with respect to the natural language learned model. The learning model obtained by learning is acquired from the other device. However, the learning support device 10 may store the natural language learned model in the storage device 14.

In the present embodiment, the learning unit 23 includes a plurality of learning units 123, each of which has a function of the learning unit 23. Each of the learning units 123 corresponds to each of the plurality of data preprocessing units 122, and the trained model is trained by training the processed script statement created by the corresponding data preprocessing unit 122 in the natural language pretrained model. Are generated respectively. However, the learning unit 23 may be realized by one learning unit, another software module, or the like.

In one preferred example, the natural language pre-learned model is a trained model called BERT provided by Google, where the learning unit 123 communicates with the BERT system via the communication device 15 to the BERT. Learn the processed script statement. In this case, the learning unit 123 uses the processed script statement of the natural language data as the training data, fine-tunes the natural language pre-trained model for each type of control data, and fine-tunes the trained model for each type of control data. To generate. Fine tuning means that the natural language pre-trained model is retrained and the parameters are reweighted. Therefore, the learning unit 123 generates a new learning model in which the natural language pre-learned model is fine-tuned by re-learning the already trained natural language pre-learned model using the processed script statement. Is.

In the present embodiment, the learning unit 123 causes the natural language pre-learned model to learn the relationship between sentences. The data preprocessing unit 122 creates a processed script sentence from the created game script shown in FIG. 5 so that the learning unit 123 can learn the relationship between sentences. FIG. 8 is a diagram showing an example of a processed script statement created by the data preprocessing unit 122 from the created game script shown in FIG. In the processed script statement, the label 36 is associated with the description 33'and the control description 35'. The processed script statement does not include the label 36, but for convenience of explanation, in FIG. 8, the label 36 given by the learning unit 23 to train the natural language pre-learned model is clearly shown. Further, for convenience of explanation, in the processed script statement shown in FIG. 8, the explanatory statement 33'and the control explanatory statement 35'are associated with the ID for the purpose of corresponding to the ID 31 of the game script, but the data of each line is managed. Anything that is possible will do.

On the label 36, "IsNext" is described as the correct answer label, and "NotNext" is described as the incorrect answer label. As a set of correct answers, the data preprocessing unit 122 has a description 33'corresponding to the character name 32 and description 33 of ID 31 "1" of the created game script shown in FIG. 5, and a command of "1" of ID 31. Create a pair with the control description 35'corresponding to the data 35. The data preprocessing unit 122 has a set of incorrect answers, a description 33'corresponding to the character name 32 and the description 33 of "1" of the ID 31 of the created game script shown in FIG. 5, and a control description randomly selected. Create a pair with sentence 35'. The learning unit 123 associates the correct answer set with the "IsNext" label 36 and the incorrect answer set with the "NotNext" label 36.

The learning unit 123 concatenates the explanatory text 33'and the control explanatory text 35'created as a set of correct answers to the natural language pre-learned model and trains them as correct answer data (with a correct answer label). , The explanation sentence 33'and the control explanation sentence 35'created as a set of incorrect answers are concatenated and learned as incorrect answer data (with an incorrect answer label). As a modification, when command1 and command2 are one type of control data and the control data is one type of control data to be learned, the data preprocessing unit 122 has a control corresponding to each of command1 and command2. It is configured so that the description 35'can be created. In this case, the learning unit 123 concatenates the control explanations 35'corresponding to each of command1 and command2 and treats them as one control explanation 35A'. The relationship between the sentences of the control explanation sentence 35A'is learned.

Next, the operations of the data preprocessing unit 22 and the learning unit 23 when the type of control data to be learned is the image path 34 will be described mainly with respect to the differences from the case of the command data 35. The image path 34 includes a file name composed of a character string including a character ID and an emotion. For example, when the character ID is "3050", if the file name is "3050", the emotion is "neutral", if it is "3050_lauch", the emotion is "laughing", if it is "3050_angry", the emotion is "angry", and "3050_concentration". If it is, it becomes the feeling of "other", and if it is "3050_a", it becomes the feeling of "other". In this case, the processed script statement shown in FIG. 6 includes the control explanation statement 34'corresponding to the emotion included in the character string of the image path 34, not the control explanation statement 35'corresponding to the command data 35. Further, in this case, since the control data depends on the individuality of the character, the data division unit 21 divides the created game script for each character.

FIG. 9 is a diagram showing a correspondence relationship between the character string included in the image path 34 and emotions, and is a diagram showing an example of a conversion table showing the correspondence between the image path 34 and the control explanatory text 34'. The data preprocessing unit 122 converts the emotion included in the character string of the image path 34 into the control description 34'"emotion" by using the conversion table shown in FIG. The data preprocessing unit 122 converts the control data included in the created game script stored in the storage device 14 for each character into a control description, and creates a processed script statement for each character. In this case, the processed script statement is the "emotion" of the control explanatory statement 34'corresponding to the image path 34 instead of the control explanatory statement 35'corresponding to the command data 35 in the processed script statement shown in FIG. 6 or FIG. ”Is included. The data preprocessing unit 122 creates a set of the explanatory text 33'and the control explanatory text 34'created based on the created game script as a set of correct answers in the same manner as in the case of the control data of command1. Similarly, the data preprocessing unit 122 creates a set of a description 33'created based on the created game script and a randomly selected control description 34' as a set of incorrect answers.

Next, the process of generating the trained model of the learning support device 10 according to the embodiment of the present invention will be described with reference to the flowchart shown in FIG. In step 101, the data division unit 21 classifies the created game scripts for each character and stores the created game scripts for each character in the storage device 14.

In step 102, the data preprocessing unit 22 converts the control data included in the created game script stored for each character into a control description for each type of control data, and converts the processed script statement for each character. create. In step 103, the learning unit 23 generates a trained model for each character by training the processed script statement created for each character in the natural language pre-learned model for each type of control data.

However, if the created game script is not divided by the data division unit 21, this flowchart does not execute the process of step 101. In this case, in step 102, the data preprocessing unit 22 converts the control data included in the stored created game script into a control description for each type of control data, and creates a processed script statement. In step 103, the learning unit 23 generates a learned model by training the created processed script statement in the natural language pre-learned model for each type of control data.

FIG. 11 is a block diagram showing a hardware configuration of the generation support device 50 according to the second embodiment of the present invention. The generation support device 50 includes a processor 51, an input device 52, a display device 53, a storage device 54, and a communication device 55. Each of these components is connected by a bus 56. It is assumed that an interface is interposed between the bus 56 and each component device as needed. The generation support device 50 includes a configuration similar to that of a general server, PC, or the like.

The processor 51 controls the operation of the entire generation support device 50. For example, the processor 51 is a CPU. The processor 51 executes various processes by reading and executing a program or data stored in the storage device 54. The processor 51 may be composed of a plurality of processors.

The input device 52 is a user interface that receives input from the user to the generation support device 50, and is, for example, a touch panel, a touch pad, or a button. The display device 53 is a display that displays an application screen or the like to the user of the generation support device 50 under the control of the processor 51.

The storage device 54 includes a main storage device and an auxiliary storage device. The main storage device is a semiconductor memory such as a RAM. The RAM is a volatile storage medium capable of reading and writing information at high speed, and is used as a storage area and a work area when the processor 51 processes information. The main storage device may include a ROM, which is a read-only non-volatile storage medium. The auxiliary storage device stores various programs and data used by the processor 51 when executing each program. The auxiliary storage device may be any non-volatile storage or non-volatile memory as long as it can store information, and may be removable.

The storage device 54 stores the learned model for each type of control data generated by the learning unit 23 of the learning support device 10. When a trained model is generated for each character according to the type of control data, the storage device 54 stores the trained model for each type of control data and for each character. Further, the storage device 14 stores a character image that is referred to when the game script is read. As described above, the file name of the character image is composed of a character string including a character ID and emotions.

The communication device 55 is a wireless LAN module capable of exchanging data with another computer such as a user terminal or a server via a network. The communication device 55 can be another wireless communication device such as a Bluetooth® module, or a wired communication device such as an Ethernet® module or a USB interface.

FIG. 12 is a functional block diagram of the generation support device 50 according to the embodiment of the present invention. The generation support device 50 includes an input receiving unit 61, a data processing unit 62, an inference unit 63, and a data post-processing unit 64. In the present embodiment, these functions are realized by executing the program by the processor 51. For example, the program to be executed is a program stored in the storage device 54 or received via the communication device 55. In this way, since various functions are realized by reading the program, a part or all of one part (function) may be possessed by another part. However, these functions may be realized by hardware by configuring an electronic circuit or the like for realizing a part or all of each function.

The input receiving unit 61 accepts the input of the character name 32 and the description 33 of the game script. In one example, the input receiving unit 61 receives the input of the character name 32 and the explanatory text 33 in the game script shown in FIG. 3, and the received data is associated with the ID 31. However, when the type of control data to be inferred is command 1, the control data does not depend on the individuality of the character, so the input receiving unit 61 may be configured to receive only the explanatory text 33.

The data processing unit 62 selects the natural language data of one explanatory text 33'as shown in Table 1 based on the natural language data of the character name 32 and the explanatory text 33, one by one in the row associated with the ID 31. create.

The inference unit 63 uses a trained model for each type of control data generated by the learning unit 23 of the learning support device 10, or a trained model for each type of control data and for each character, and the input reception unit 61 uses the trained model. The control explanation is inferred from the explanation that received the input. The data post-processing unit 64 creates control data from the control explanatory text inferred by the inference unit 63 based on the control data conversion information. Specifically, when the type of control data is command data 35, the data post-processing unit 64 converts the control description into control data based on the control data conversion information.

In one example, when the type of control data to be inferred is command 1, the inference unit 63 has a control description corresponding to the explanatory text 33'created by the data processing unit 62 and all types of command data described in command 1. Create a pair with each of sentence 35'. The inference unit 63 inputs the created set to the trained model corresponding to command1 and calculates the relevance score. The inference unit 63 selects (outputs) the control explanation sentence 35'of the set with the highest score as the most appropriate control explanation sentence.

FIG. 13 is a diagram showing an example of estimation of the control explanatory text 35'by the inference unit 63, and Table 2 shows the explanatory text 33'created by the data processing unit 62.

Since the current speaker character is "Bee" and the immediately preceding speaker character is "Rheinhalza", the data processing unit 62 adds "Characters are Rheinhalza, Bie " to the dialogue in the explanation 33. Then, the explanation 33'was created. The inference unit 63 has created a description 33'and a control description corresponding to all types of command data "Nothing happened", "(Character) appeared", "(Character) disappeared", "(Character) disappeared", " Create a pair with each of "(character) jumped" and "something happened". The control explanations of all types correspond to the command data "NoCommand", "fadein", "fadeout", "jump", and "other" of all types. The inference unit 63 inputs the created set to the trained model corresponding to command1 to calculate the relevance score, and gives the highest score "(character) jumped" to the explanation 33. Select as the most appropriate control description.

The data post-processing unit 64 converts the control explanatory text inferred by the inference unit by the inference unit 63 into control data, and the input receiving unit 61 receives the input and the control explanatory text corresponding to the explanatory text 33. Output sentence 35'. In one preferred example, the storage device 54 stores the control data conversion information for each type of control data stored in the storage device 14 of the learning support device 10. The data post-processing unit 64 converts (creates) the control description into control data based on the control data conversion information. The data post-processing unit 64 is configured to output the converted control data to the corresponding game script location.

In one example, when the type of control data to be inferred is the image path 34, the inference unit 63 corresponds to the explanatory text 33'created by the data processing unit 62 and all types of emotions included in the image path 34. Create a pair with each of the control description 34'. The inference unit 63 inputs the created set to the trained model corresponding to the character corresponding to the character name 32 received by the input receiving unit 61 and corresponding to the image path 34, and calculates the relevance score. .. The inference unit 63 inputs the created set to the trained model corresponding to the image path 34, calculates the relevance score, and sets the control explanatory text 34'with the highest score to the explanatory text 33 most. Select as the appropriate control description. For example, the data post-processing unit 64 converts the control description into a character string including emotions such as “laugh” and “angry” based on the control data conversion information. The data post-processing unit 64 creates an image path (control data) in which the character ID specified from the description 33'and the file name including the converted character string are stored. The data post-processing unit 64 is configured to output the created image path to the corresponding game script location. In another example, since the control explanation 34'is natural language data corresponding to emotions, the generation support device 50 does not process the control explanation 34' with the highest score by the data post-processing unit 64. , It is configured so that it can be output in the form of a comment to the corresponding game script as it is.

Next, a part of the automatic creation process of the game script of the generation support device 50 according to the embodiment of the present invention will be described with reference to the flowchart shown in FIG. In step 201, the input receiving unit 61 accepts the input of the character name 32 and the explanatory text 33. In step 201, the data processing unit 62 creates the explanatory text 33'as shown in Table 1 based on the natural language data of the character name 32 and the explanatory text 33.

In step 202, the inference unit 63 uses the learned model for each type of control data generated by the learning unit 23 of the learning support device 10 to generate a control explanation from the explanation that the input reception unit 61 has received the input. Infer. Specifically, the inference unit 63 outputs one control explanation. In step 203, the data post-processing unit 64 converts the control explanatory text inferred (output) by the inference unit by the inference unit 63 into control data, and the explanatory text 33 and the explanation in which the input receiving unit 61 receives the input. The control explanation corresponding to the sentence 33 is output.

Next, the main actions and effects of the learning support device 10 and the generation support device 50 according to the embodiment of the present invention will be described. In the embodiment of the present invention, the learning support device 10 includes a data preprocessing unit 22 and a learning unit 23. In order to use the created game script as learning data, the data preprocessing unit 22 converts the control data for training into a control explanation and makes it a natural language to create a processed script statement. The data preprocessing unit 22 is configured to be able to create a processed script statement for each type of control data and for each character. The learning unit 23 generates a trained model by training a processed script statement in a natural language pre-learned model. The learning unit 23 is configured to be able to generate a trained model for each type of control data and for each character by using a processed script statement created for each type of control data and for each character.

As described above, in the present embodiment, the learning support device 10 makes all the data to be learned into natural language by converting the control data into natural language, and uses the natural languageized learning data to make a BERT or the like. A trained model is generated by retraining a natural language pre-trained model.

With such a configuration, in the present embodiment, it is possible to generate a trained model for realizing automatic creation of at least a part of the game script by using the created game script.
Since the trained model can be generated for each type of control data and also for each character, an independent trained model is generated according to the type of control data to be automatically created and the type of character. It becomes possible to do. That is, it is possible to generate different trained models according to a plurality of characters and estimation targets (types of control data). This makes it possible to generate a trained model that can be applied to a plurality of narrative games with different characters and control data.

The generation support device 50 includes an inference unit 63 and a data post-processing unit 64. The inference unit 63 is configured to be able to infer and output a control explanation for each type of control data when, for example, receiving an input of dialogue using the learned model generated by the learning support device 10. The inference unit 63 is configured to be able to infer and output a control explanation for each character according to the type of control data. The data post-processing unit 64 is configured to be able to convert the control explanatory text, which is the natural language data output by the inference unit 63, into the control data by using the control data conversion information. The data post-processing unit 64 is configured to be able to output the converted control data to the corresponding game script location.

With such a configuration, in the present embodiment, when a character name or a dialogue is input to the game script, control data can be automatically output. In this way, by semi-automating the game script creation work that was conventionally performed manually, it is possible to reduce the labor required for the game developer to create the game script. In this way, it is possible to support the creation of at least a part of the game script.

Further, in the present embodiment, the learning support device 10 generates learning data by combining natural language data without using tags such as XML tags. With such a configuration, it is possible to streamline the game script creation process by eliminating the need for tag convention agreements and annotations, and it is possible to reduce the labor required for game developers to create game scripts. It will be possible.

Until now, in order to learn a natural language, it is necessary to perform annotation work to add attribute information to the natural language to be learned, and in order to prepare learning data, humans read and tag all the learning data. Had to do. The learning support device 10 of the present embodiment is, for example, BERT ("Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina, BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, 2018, arXiv preprint arXiv". : 1810.04805. ") is used. As a result, learning can be performed simply by automatically adding the natural language including the attribute information required for estimation to the dialogue without tagging.

Unless otherwise specified, the above-mentioned effects are the same in other embodiments and examples.

In another embodiment of the present invention, it can be an automatic creation support device having the configurations of the learning support device 10 and the generation support device 50 described above. The automatic creation support device can be an automatic creation support system composed of a plurality of devices or the like, or can be realized as a system including a learning support device 10 and a generation support device 50.

In another embodiment of the present invention, the learning support device 10 does not include the data dividing unit 21 when all the control data does not depend on the character or when it is not necessary to train each character.

In another embodiment of the present invention, when the control data cannot be classified into a plurality of types by a function or does not need to be classified, the learning support device 10 and the generation support device 50 perform each process for each type of control data. Do not execute. For example, the data preprocessing unit 22 does not classify by the type of control data, but converts the control data included in the created game script stored in the storage device 14 into a control description which is natural language data, and processes the processed script. The sentence is created, and the learning unit 23 generates the trained model by training the processed script sentence in the natural language pre-trained model without classifying the control data according to the type. Also in the embodiment, the data division unit 21 can classify the created game scripts for each character and store the created game scripts for each character in the storage device 14. In this case, the data preprocessing unit 22 creates a processed script statement for each character, and the learning unit 23 generates a trained model for each character for each type of control data.

In another embodiment of the invention, the game script is structured data, such as a tree structure, instead of matrix data. In this case, the row element identifiers in the case of matrix-formatted data are associated with the tree node identifiers or paths, and each of the column elements is associated with each of the nodes.

In another embodiment of the present invention, a program that realizes the functions of the embodiment of the present invention described above and the information processing shown in the flowchart, or a computer-readable storage medium that stores the program can also be used. Further, in another embodiment, it is also possible to use an electronic device that independently realizes the functions of the embodiment of the present invention described above and the information processing shown in the flowchart. Further, in another embodiment, it is also possible to use a method for realizing the functions of the embodiment of the present invention described above and the information processing shown in the flowchart. In another embodiment, the server may be a server capable of supplying the computer with a program that realizes the functions of the embodiment of the present invention described above and the information processing shown in the flowchart. Further, in another embodiment, it can be a virtual machine that realizes the functions of the embodiment of the present invention described above and the information processing shown in the flowchart.

In the processing or operation described above, the processing or operation can be freely performed in a certain step as long as there is no contradiction in the processing or operation such as using data that should not be available in that step. Can be changed. Further, each of the examples described above is an example for explaining the present invention, and the present invention is not limited to these examples. The present invention can be carried out in various forms as long as it does not deviate from the gist thereof.

10 Learning support device 11 Processor 12 Input device 13 Display device 14 Storage device 15 Communication device 16 Bus 21 Data division unit 22 Data preprocessing unit 23 Learning unit 31 ID
32 Character name 33, 33'Description 34 Image path 34'Control description 35 Command data 35'Control description 36 Label 40 Game screen 41 Standing picture 42 Character name 43 Description 50 Generation support device 51 Processor 52 Input device 53 Display Device 54 Storage device 55 Communication device 56 Bus 61 Input reception unit 62 Data processing unit 63 Inference unit 64 Data post-processing unit 122 Data pre-processing unit 123 Learning unit

Claims (13)

It is a system for supporting the creation of a game script including natural language data representing a description in a game associated with the content of the game and control data for controlling the game.
Data preprocessing unit that converts the control data included in the pre-created game script into a control description that is natural language data, and creates a processed script statement that includes the description and the control description corresponding to the description. When,
A learning unit that generates a trained model by training a processed script sentence in a natural language pre-learned model in which the grammatical structure related to the natural language and the relationship between sentences are learned in advance.
Including the system. Control data can be classified into multiple types according to the function.
The data preprocessing unit converts the control data included in the pre-created game script for each type of control data into a control description which is natural language data, and the description and the control description corresponding to the description. Create a processed script statement containing the statement and
The learning unit trains a processed script sentence in a natural language pre-learned model in which the grammatical structure related to the natural language and the relationship between sentences are learned in advance for each type of control data. The system according to claim 1, which generates a trained model of. The data preprocessing unit includes a plurality of data preprocessing units, and one data preprocessing unit converts control data corresponding to one type of control data included in the created game script into a control description. By creating a processed script statement, the data preprocessing unit creates a processed script statement for each type of control data.
The system according to claim 2, wherein the learning unit includes a plurality of learning units corresponding to each of the plurality of data preprocessing units. The system according to any one of claims 1 to 3, wherein the learning unit uses a processed script statement as training data to fine-tune a natural language pre-learned model to generate a trained model. The processed script statement further includes a descriptive text and a randomly selected control descriptive text.
The learning unit causes the natural language pre-learned model to learn the explanatory text included in the processed script statement and the control explanatory text corresponding to the explanatory text as correct data, and the explanatory text and randomly selected control. The system according to any one of claims 1 to 4, wherein the explanation is trained as incorrect data to generate a trained model. From claim 1, the game script is matrix data or structured data, each of which comprises a plurality of identifiers corresponding to individual scenes of the game, and natural language data and control data associated with the identifiers. The system according to any one of 5. The game script also contains natural language data representing the character name associated with the identifier.
It further includes a data division unit that classifies created game scripts by character and stores the created game scripts for each character.
The data preprocessing unit converts the control data included in the created game script into a control explanation for each character, and creates an explanation and a processed script including the control explanation corresponding to the explanation.
The system according to claim 6, wherein the learning unit generates a learned model for each character by training a processed script sentence in a natural language pre-learned model for each character. The data preprocessing unit converts the control data into the control description based on the conversion information indicating the correspondence between the control data and the control description.
The system
An input reception unit that accepts input of explanations in the game,
The input reception unit accepted the input using the trained model generated by training the processed script sentence in the natural language pre-trained model in which the grammatical structure and the relationship between sentences related to the natural language were trained in advance. It is an inference unit that infers a control explanation from the explanation, and the processed script statement is a natural language data created from the explanation included in the created game script created in advance and the control data corresponding to the explanation. The inference part, which contains a certain control description,
The system according to any one of claims 1 to 7, which includes a data post-processing unit that creates control data from a control description inferred by the inference unit based on the conversion information. A method of generating a trained model to assist in the creation of a game script containing natural language data representing a description in a game associated with the content of the game and control data for controlling the game.
A step of converting the control data included in the pre-created game script into a control description which is natural language data, and creating a processed script statement including the description and the control description corresponding to the description.
A step to generate a trained model by training a processed script sentence in a natural language pre-learned model in which the grammatical structure related to the natural language and the relationship between sentences are learned in advance.
Including methods. It is a system for supporting the creation of a game script including natural language data representing a description in a game associated with the content of the game and control data for controlling the game.
An input reception unit that accepts input of explanations in the game,
The input reception unit accepted the input using the trained model generated by training the processed script sentence in the natural language pre-trained model in which the grammatical structure and the relationship between sentences related to the natural language were trained in advance. It is an inference unit that infers a control explanation from the explanation, and the processed script statement is a natural language data created from the explanation included in the created game script created in advance and the control data corresponding to the explanation. The inference part, which contains a certain control description,
Including the system. The system according to claim 10, further comprising a data post-processing unit that creates control data from the control explanation inferred by the inference unit based on the conversion information indicating the correspondence between the control data and the control explanation. A method for supporting the creation of a game script containing natural language data representing a description in a game and control data for controlling the game.
Steps to accept input of description in the game,
Control from the explanatory text that accepts the input using the trained model generated by training the processed script sentence in the natural language pre-trained model in which the grammatical structure and the relationship between sentences related to the natural language are trained in advance. The processed script statement is a step of inferring the explanatory statement, and the processed script statement is a control explanatory statement that is natural language data created from the explanatory statement included in the created game script created in advance and the control data corresponding to the explanatory statement. Including, steps and
Including methods. A program that causes a computer to perform each step of the method according to claim 9 or 12.

Images